The "iq" selection (test) - update

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Rebel, chrisw

lech
Posts: 1136
Joined: Sun Feb 14, 2010 10:02 pm

Re: The "iq" selection (test) - update

Post by lech »

Plutie wrote: Thu Jan 26, 2023 5:06 pm #30 also has Nc3+ - if Kc1, then Kf8 wins, if Ke1, then Re2+, Kf1, and Ra2 wins.

[pgn][Event "?"]
[Site "?"]
[Date "2023.01.26"]
[Round "?"]
[White "Lc0 v0.30.0-dag+git.dirty"]
[Black "Lc0 v0.30.0-dag+git.dirty"]
[Result "*"]
[FEN "q1b4n/1p1pKp2/1p1p1Pp1/1P1P2Pp/2P4P/8/7R/1N1k4 w - - 0 1"]
[SetUp "1"]

1. Nc3+ {[%clk 00:02:51]} 1... Kc1 {[%clk 00:02:51]} 2. Kf8 {[%clk 00:02:53]}
2... Qa3 {[%clk 00:02:50]} 3. Ne2+ {[%clk 00:02:53]} 3... Kd2 {[%clk 00:02:37]}
4. Kg8 {[%clk 00:02:53]} 4... Qa1 {[%clk 00:02:28]} 5. Kxh8 {[%clk 00:02:55]}
5... Qe5 {[%clk 00:02:25]} 6. Nf4+ {[%clk 00:02:56]} 6... Kc3 {[%clk 00:02:22]}
7. Rf2 {[%clk 00:02:57]} 7... Qe1 {[%clk 00:02:20]} 8. Rf3+ {[%clk 00:02:56]}
8... Kxc4 {[%clk 00:02:18]} 9. Kg7 {[%clk 00:02:49]} 9... Qh1 {[%clk 00:01:59]}
10. Rh3 {[%clk 00:02:50]} 10... Qf1 {[%clk 00:01:54]} 11. Kxf7 {[%clk
00:02:52]} 11... Qxf4 {[%clk 00:01:55]} 12. Kg8 {[%clk 00:02:53]} 12... Qf5
{[%clk 00:01:54]} 13. Rh2 {[%clk 00:02:47]} 13... Qxd5+ {[%clk 00:01:35]} 14.
f7 {[%clk 00:02:48]} 14... Kxb5 {[%clk 00:01:35]} 15. Rc2 {[%clk 00:02:50]}
15... Qc6 {[%clk 00:01:34]} 16. Rb2+ {[%clk 00:02:43]} 16... Ka6 {[%clk
00:01:28]} 17. f8=Q {[%clk 00:02:44]} 17... Ka7 {[%clk 00:01:20]} 18. Kh7
{[%clk 00:02:32]} 18... d5 {[%clk 00:01:20]} 19. Qf2 {[%clk 00:02:23]} 19... d6
{[%clk 00:01:21]} 20. Kxg6 {[%clk 00:02:22]} 20... Qe8+ {[%clk 00:01:17]} 21.
Kg7 {[%clk 00:02:11]} 21... Qe5+ {[%clk 00:01:11]} 22. Qf6 {[%clk 00:02:13]}
22... Bf5 {[%clk 00:01:08]} *

[/pgn]
If you are right (it is possible). to avoid it, White Rook can start from g2 instead of h2.
Maybe, I can't be friendly, but let me be useful.
peter
Posts: 3206
Joined: Sat Feb 16, 2008 7:38 am
Full name: Peter Martan

Re: The "iq" selection (test) - update

Post by peter »

lech wrote: Sat Jan 28, 2023 12:12 am
Plutie wrote: Thu Jan 26, 2023 5:06 pm #30 also has Nc3+ - if Kc1, then Kf8 wins, if Ke1, then Re2+, Kf1, and Ra2 wins.

[pgn][Event "?"]
[Site "?"]
[Date "2023.01.26"]
[Round "?"]
[White "Lc0 v0.30.0-dag+git.dirty"]
[Black "Lc0 v0.30.0-dag+git.dirty"]
[Result "*"]
[FEN "q1b4n/1p1pKp2/1p1p1Pp1/1P1P2Pp/2P4P/8/7R/1N1k4 w - - 0 1"]
[SetUp "1"]

1. Nc3+ {[%clk 00:02:51]} 1... Kc1 {[%clk 00:02:51]} 2. Kf8 {[%clk 00:02:53]}
2... Qa3 {[%clk 00:02:50]} 3. Ne2+ {[%clk 00:02:53]} 3... Kd2 {[%clk 00:02:37]}
4. Kg8 {[%clk 00:02:53]} 4... Qa1 {[%clk 00:02:28]} 5. Kxh8 {[%clk 00:02:55]}
5... Qe5 {[%clk 00:02:25]} 6. Nf4+ {[%clk 00:02:56]} 6... Kc3 {[%clk 00:02:22]}
7. Rf2 {[%clk 00:02:57]} 7... Qe1 {[%clk 00:02:20]} 8. Rf3+ {[%clk 00:02:56]}
8... Kxc4 {[%clk 00:02:18]} 9. Kg7 {[%clk 00:02:49]} 9... Qh1 {[%clk 00:01:59]}
10. Rh3 {[%clk 00:02:50]} 10... Qf1 {[%clk 00:01:54]} 11. Kxf7 {[%clk
00:02:52]} 11... Qxf4 {[%clk 00:01:55]} 12. Kg8 {[%clk 00:02:53]} 12... Qf5
{[%clk 00:01:54]} 13. Rh2 {[%clk 00:02:47]} 13... Qxd5+ {[%clk 00:01:35]} 14.
f7 {[%clk 00:02:48]} 14... Kxb5 {[%clk 00:01:35]} 15. Rc2 {[%clk 00:02:50]}
15... Qc6 {[%clk 00:01:34]} 16. Rb2+ {[%clk 00:02:43]} 16... Ka6 {[%clk
00:01:28]} 17. f8=Q {[%clk 00:02:44]} 17... Ka7 {[%clk 00:01:20]} 18. Kh7
{[%clk 00:02:32]} 18... d5 {[%clk 00:01:20]} 19. Qf2 {[%clk 00:02:23]} 19... d6
{[%clk 00:01:21]} 20. Kxg6 {[%clk 00:02:22]} 20... Qe8+ {[%clk 00:01:17]} 21.
Kg7 {[%clk 00:02:11]} 21... Qe5+ {[%clk 00:01:11]} 22. Qf6 {[%clk 00:02:13]}
22... Bf5 {[%clk 00:01:08]} *

[/pgn]
If you are right (it is possible). to avoid it, White Rook can start from g2 instead of h2.
Plutie is right, the study by Eisert & Rehm in that way is somewhat cooked by a major dual, but Sting should accept this second one solutioin from original position too, shouldn't it?

Crystal after some Forward- Backward of both winning lines in depth 36 MultiPV=2

q1b4n/1p1pKp2/1p1p1Pp1/1P1P2Pp/2P4P/8/7R/1N1k4 w - - 0 1

Analysis by Crystal 5 KWK:

1. +- (19.18): 1.Ra2 Qxa2 2.Nc3+ Kc2 3.Nxa2 Kb3 4.Kf8 Kxc4 5.Kg7 Kxd5 6.Kxh8 Ke6 7.Nc3 d5 8.Ne2 d6 9.Kg8 Ke5 10.Kxf7 Bf5 11.Kg7 d4 12.f7 d3 13.Nc3 Kd4 14.Nd1 Bg4 15.Nf2 Ke3 16.f8Q Bf5 17.Nxd3 Kxd3 18.Qxd6+ Kc4 19.Qxb6 Be4 20.Kf6 Kc3 21.Qe3+ Bd3 22.Qc5+ Kb3 23.Qd5+ Bc4 24.Qf3+ Ka4
2. +- (7.61): 1.Nc3+ Kc1 2.Kf8 Qa5 3.Ne2+ Kd2 4.Kg7 Qa4 5.Kxh8 Qxc4 6.Kg7 Kd1 7.Kxf7 Qxd5+ 8.Kg7 Qe5 9.Rf2 Qe3 10.Rg2 Qf3 11.Rg1+ Kxe2 12.Rc1 d5 13.Rxc8 d4 14.Re8+ Kf2 15.f7 d3 16.Rd8 d2 17.Rxd7 Kg3 18.f8Q d1Q 19.Rxd1 Qxd1 20.Kxg6 Qd2 21.Kh7 Qd7+ 22.Qg7 Qxb5 23.g6 Qf5 24.Qe7 Qc2 25.Qf6 Kh2 26.Kh6 Qc1+ 27.Qg5 Qc6 28.Qf5 b5 29.Kh7 b4 30.g7 Qc7 31.Qf6

Instead you have to go into the side line up to 5 moves deep with Sting Black Hole 4 (the crucial 2.Kf8! after 1.Nc3+ Kc1 isn't enough, not until after 3.Ne2+ 4.Kg8 5.Kxh8 the engine starts raising its eval) to make it see the win from there onwards too.

Yet you are right too, R at g2 instead of h2 should work for a repair to make the major dual impossible:

q1b4n/1p1pKp2/1p1p1Pp1/1P1P2Pp/2P4P/8/6R1/1N1k4 w - - 0 1

Analysis by Crystal 5 KWK:

1. +- (13.41): 1.Ra2 Qxa2 2.Nc3+ Ke1 3.Nxa2 Kf2 4.Nc3 Kg3 5.Ne4+ Kxh4 6.Nxd6 Kxg5 7.Nxf7+ Nxf7 8.Kxf7 h4 9.Ke7 h3 10.f7 h2 11.f8Q h1Q 12.Qxc8 Qh7+ 13.Kd8 Kf5 14.Qxd7+ Qxd7+ 15.Kxd7 g5 16.d6 g4 17.Kc8 g3 18.d7 g2 19.d8Q g1Q 20.Kxb7 Qe3 21.Qxb6 Qd3 22.Qc6 Kg4 23.c5 Kh3 24.Qe6+ Kg3 25.Qe5+ Kg2 26.Qg5+ Kf1 27.c6 Qf3 28.Qc5 Qg2 29.Qf8+ Ke2 30.Qe8+ Kd3 31.Qe6 Qh1

2. = (0.00): 1.Nc3+ Kc1 2.Na2+ Kd1

And if now 1.Nc3+ Kc1 2.Kf8:

q1b2K1n/1p1p1p2/1p1p1Pp1/1P1P2Pp/2P4P/2N5/6R1/2k5 b - - 0 1

Analysis by Crystal 5 KWK:

2...Qa3 3.Ne2+ Kd1 4.Kg7 Qh3 5.Nf4 Qe3 6.Ne2
= (0.00) Depth: 49/10 00:00:11 324MN
Peter.
lech
Posts: 1136
Joined: Sun Feb 14, 2010 10:02 pm

Re: The "iq" selection (test) - update

Post by lech »

peter wrote: Sat Jan 28, 2023 12:32 pm
lech wrote: Sat Jan 28, 2023 12:12 am
Plutie wrote: Thu Jan 26, 2023 5:06 pm #30 also has Nc3+ - if Kc1, then Kf8 wins, if Ke1, then Re2+, Kf1, and Ra2 wins.

[pgn][Event "?"]
[Site "?"]
[Date "2023.01.26"]
[Round "?"]
[White "Lc0 v0.30.0-dag+git.dirty"]
[Black "Lc0 v0.30.0-dag+git.dirty"]
[Result "*"]
[FEN "q1b4n/1p1pKp2/1p1p1Pp1/1P1P2Pp/2P4P/8/7R/1N1k4 w - - 0 1"]
[SetUp "1"]

1. Nc3+ {[%clk 00:02:51]} 1... Kc1 {[%clk 00:02:51]} 2. Kf8 {[%clk 00:02:53]}
2... Qa3 {[%clk 00:02:50]} 3. Ne2+ {[%clk 00:02:53]} 3... Kd2 {[%clk 00:02:37]}
4. Kg8 {[%clk 00:02:53]} 4... Qa1 {[%clk 00:02:28]} 5. Kxh8 {[%clk 00:02:55]}
5... Qe5 {[%clk 00:02:25]} 6. Nf4+ {[%clk 00:02:56]} 6... Kc3 {[%clk 00:02:22]}
7. Rf2 {[%clk 00:02:57]} 7... Qe1 {[%clk 00:02:20]} 8. Rf3+ {[%clk 00:02:56]}
8... Kxc4 {[%clk 00:02:18]} 9. Kg7 {[%clk 00:02:49]} 9... Qh1 {[%clk 00:01:59]}
10. Rh3 {[%clk 00:02:50]} 10... Qf1 {[%clk 00:01:54]} 11. Kxf7 {[%clk
00:02:52]} 11... Qxf4 {[%clk 00:01:55]} 12. Kg8 {[%clk 00:02:53]} 12... Qf5
{[%clk 00:01:54]} 13. Rh2 {[%clk 00:02:47]} 13... Qxd5+ {[%clk 00:01:35]} 14.
f7 {[%clk 00:02:48]} 14... Kxb5 {[%clk 00:01:35]} 15. Rc2 {[%clk 00:02:50]}
15... Qc6 {[%clk 00:01:34]} 16. Rb2+ {[%clk 00:02:43]} 16... Ka6 {[%clk
00:01:28]} 17. f8=Q {[%clk 00:02:44]} 17... Ka7 {[%clk 00:01:20]} 18. Kh7
{[%clk 00:02:32]} 18... d5 {[%clk 00:01:20]} 19. Qf2 {[%clk 00:02:23]} 19... d6
{[%clk 00:01:21]} 20. Kxg6 {[%clk 00:02:22]} 20... Qe8+ {[%clk 00:01:17]} 21.
Kg7 {[%clk 00:02:11]} 21... Qe5+ {[%clk 00:01:11]} 22. Qf6 {[%clk 00:02:13]}
22... Bf5 {[%clk 00:01:08]} *

[/pgn]
If you are right (it is possible). to avoid it, White Rook can start from g2 instead of h2.
Plutie is right, the study by Eisert & Rehm in that way is somewhat cooked by a major dual, but Sting should accept this second one solutioin from original position too, shouldn't it?
...
Of course it is not necessary to see any engine (Sting too) solving this strange position (after 1. Nc3+). This side (cook) solution is, in my opinion, very complex and ugly and worthless as study (puzzle).
The main surprising idea (after 1. Ra2) is wise (and beautiful) and each AI engine should find it.
Maybe, I can't be friendly, but let me be useful.
lech
Posts: 1136
Joined: Sun Feb 14, 2010 10:02 pm

Re: The "iq" selection (test) - update

Post by lech »

peter wrote: Sat Jan 28, 2023 12:32 pm
lech wrote: Sat Jan 28, 2023 12:12 am
Plutie wrote: Thu Jan 26, 2023 5:06 pm #30 also has Nc3+ - if Kc1, then Kf8 wins, if Ke1, then Re2+, Kf1, and Ra2 wins.

[pgn][Event "?"]
[Site "?"]
[Date "2023.01.26"]
[Round "?"]
[White "Lc0 v0.30.0-dag+git.dirty"]
[Black "Lc0 v0.30.0-dag+git.dirty"]
[Result "*"]
[FEN "q1b4n/1p1pKp2/1p1p1Pp1/1P1P2Pp/2P4P/8/7R/1N1k4 w - - 0 1"]
[SetUp "1"]

1. Nc3+ {[%clk 00:02:51]} 1... Kc1 {[%clk 00:02:51]} 2. Kf8 {[%clk 00:02:53]}
2... Qa3 {[%clk 00:02:50]} 3. Ne2+ {[%clk 00:02:53]} 3... Kd2 {[%clk 00:02:37]}
4. Kg8 {[%clk 00:02:53]} 4... Qa1 {[%clk 00:02:28]} 5. Kxh8 {[%clk 00:02:55]}
5... Qe5 {[%clk 00:02:25]} 6. Nf4+ {[%clk 00:02:56]} 6... Kc3 {[%clk 00:02:22]}
7. Rf2 {[%clk 00:02:57]} 7... Qe1 {[%clk 00:02:20]} 8. Rf3+ {[%clk 00:02:56]}
8... Kxc4 {[%clk 00:02:18]} 9. Kg7 {[%clk 00:02:49]} 9... Qh1 {[%clk 00:01:59]}
10. Rh3 {[%clk 00:02:50]} 10... Qf1 {[%clk 00:01:54]} 11. Kxf7 {[%clk
00:02:52]} 11... Qxf4 {[%clk 00:01:55]} 12. Kg8 {[%clk 00:02:53]} 12... Qf5
{[%clk 00:01:54]} 13. Rh2 {[%clk 00:02:47]} 13... Qxd5+ {[%clk 00:01:35]} 14.
f7 {[%clk 00:02:48]} 14... Kxb5 {[%clk 00:01:35]} 15. Rc2 {[%clk 00:02:50]}
15... Qc6 {[%clk 00:01:34]} 16. Rb2+ {[%clk 00:02:43]} 16... Ka6 {[%clk
00:01:28]} 17. f8=Q {[%clk 00:02:44]} 17... Ka7 {[%clk 00:01:20]} 18. Kh7
{[%clk 00:02:32]} 18... d5 {[%clk 00:01:20]} 19. Qf2 {[%clk 00:02:23]} 19... d6
{[%clk 00:01:21]} 20. Kxg6 {[%clk 00:02:22]} 20... Qe8+ {[%clk 00:01:17]} 21.
Kg7 {[%clk 00:02:11]} 21... Qe5+ {[%clk 00:01:11]} 22. Qf6 {[%clk 00:02:13]}
22... Bf5 {[%clk 00:01:08]} *

[/pgn]
If you are right (it is possible). to avoid it, White Rook can start from g2 instead of h2.
Plutie is right, the study by Eisert & Rehm in that way is somewhat cooked by a major dual, but Sting should accept this second one solutioin from original position too, shouldn't it?

Crystal after some Forward- Backward of both winning lines in depth 36 MultiPV=2
...
With my interference (forward-backward) each engine will solve each positions.
Don't help, AI machines should work independently of us!
Some problems, for engines, with the position after 1.Nc3+ Kc1 to find the very complex solution 2. Kf8 seems to be technical mainly. I will take a look at this later. :)
Maybe, I can't be friendly, but let me be useful.
lech
Posts: 1136
Joined: Sun Feb 14, 2010 10:02 pm

Re: The "iq" selection (test) - update

Post by lech »

peter wrote: Sat Jan 28, 2023 12:32 pm
lech wrote: Sat Jan 28, 2023 12:12 am
Plutie wrote: Thu Jan 26, 2023 5:06 pm #30 also has Nc3+ - if Kc1, then Kf8 wins, if Ke1, then Re2+, Kf1, and Ra2 wins.

[pgn][Event "?"]
[Site "?"]
[Date "2023.01.26"]
[Round "?"]
[White "Lc0 v0.30.0-dag+git.dirty"]
[Black "Lc0 v0.30.0-dag+git.dirty"]
[Result "*"]
[FEN "q1b4n/1p1pKp2/1p1p1Pp1/1P1P2Pp/2P4P/8/7R/1N1k4 w - - 0 1"]
[SetUp "1"]

1. Nc3+ {[%clk 00:02:51]} 1... Kc1 {[%clk 00:02:51]} 2. Kf8 {[%clk 00:02:53]}
2... Qa3 {[%clk 00:02:50]} 3. Ne2+ {[%clk 00:02:53]} 3... Kd2 {[%clk 00:02:37]}
4. Kg8 {[%clk 00:02:53]} 4... Qa1 {[%clk 00:02:28]} 5. Kxh8 {[%clk 00:02:55]}
5... Qe5 {[%clk 00:02:25]} 6. Nf4+ {[%clk 00:02:56]} 6... Kc3 {[%clk 00:02:22]}
7. Rf2 {[%clk 00:02:57]} 7... Qe1 {[%clk 00:02:20]} 8. Rf3+ {[%clk 00:02:56]}
8... Kxc4 {[%clk 00:02:18]} 9. Kg7 {[%clk 00:02:49]} 9... Qh1 {[%clk 00:01:59]}
10. Rh3 {[%clk 00:02:50]} 10... Qf1 {[%clk 00:01:54]} 11. Kxf7 {[%clk
00:02:52]} 11... Qxf4 {[%clk 00:01:55]} 12. Kg8 {[%clk 00:02:53]} 12... Qf5
{[%clk 00:01:54]} 13. Rh2 {[%clk 00:02:47]} 13... Qxd5+ {[%clk 00:01:35]} 14.
f7 {[%clk 00:02:48]} 14... Kxb5 {[%clk 00:01:35]} 15. Rc2 {[%clk 00:02:50]}
15... Qc6 {[%clk 00:01:34]} 16. Rb2+ {[%clk 00:02:43]} 16... Ka6 {[%clk
00:01:28]} 17. f8=Q {[%clk 00:02:44]} 17... Ka7 {[%clk 00:01:20]} 18. Kh7
{[%clk 00:02:32]} 18... d5 {[%clk 00:01:20]} 19. Qf2 {[%clk 00:02:23]} 19... d6
{[%clk 00:01:21]} 20. Kxg6 {[%clk 00:02:22]} 20... Qe8+ {[%clk 00:01:17]} 21.
Kg7 {[%clk 00:02:11]} 21... Qe5+ {[%clk 00:01:11]} 22. Qf6 {[%clk 00:02:13]}
22... Bf5 {[%clk 00:01:08]} *

[/pgn]
If you are right (it is possible). to avoid it, White Rook can start from g2 instead of h2.
Plutie is right, the study by Eisert & Rehm in that way is somewhat cooked by a major dual, but Sting should accept this second one solutioin from original position too, shouldn't it?
...
It is, in fact, a technical strange problem. It happens very rare. I saw it earlier only in one position but ignored it. When PV line is ended by 3 fold repetitions very quickly, it makes to find a very long (complex) solution simply impossible. Because there is the next position, I will try to correct it in the next version of Sting. Thank you, Plutie and Peter. :)
Maybe, I can't be friendly, but let me be useful.
peter
Posts: 3206
Joined: Sat Feb 16, 2008 7:38 am
Full name: Peter Martan

Re: The "iq" selection (test) - update

Post by peter »

lech wrote: Tue Jan 31, 2023 2:41 pm It is, in fact, a technical strange problem. It happens very rare. I saw it earlier only in one position but ignored it. When PV line is ended by 3 fold repetitions very quickly, it makes to find a very long (complex) solution simply impossible. Because there is the next position, I will try to correct it in the next version of Sting. Thank you, Plutie and Peter. :)
You're very welcome, Marek!
Peter.
electricmaster23
Posts: 2
Joined: Thu Mar 24, 2022 1:35 pm
Full name: Ben Schultz

Re: The "iq" selection (test) - update

Post by electricmaster23 »

lech wrote: Sat Jan 14, 2023 3:37 pm The authors of studies:

by YACPDB:

Mendheim Julius - 1
Blathy Otto - 3, 12, 13
Benko Pal - 16, 75
Hasek Josef - 22, 28
Gurvich Abram - 23
Kasparyan Genrikh - 24
Neghina Mihai - 27
Grasemann Herbert - 37
Rudolph William - 46
Troitzky Aleksei - 57
Reti Richard - 58
Matous Mario - 60
Behting Carl - 62
Klyatskin Mikhail - 66
Kubbel Karl - 69
Novomesky Daniel - 73
Lazard Frederic - 74
Marwitz Jan - 79
Lamford Paul - 82
Simkhovich Froim - 86

by other sources:

Solovyov Y. - 6
Babic M. - 10
Neghina Mihai - 11, 31, 34, 35 (based on study), 84, 87, 89, 92
Simkhovich Froim - 20 (part of study)
Ganguly Surya - 33
van Breukelen Gijs - 36
Hoch Yehuda & Aloni Hillel - 55
Troitzky Aleksei - 59
Smyslov Vasily - 67
Bondarenko F. & Kuznetsov A. - 68
Illescas Miguel - 72
Zemlyanski Yuri - 81 (part of study)
Stavrietsky Aleksandr - 88
Kazantsev A. - 93
Ryabinin Nikolaj - 96
Neghina Mihai & Rusz Arpad- 97
Kamody Mario & Neghina Mihai - 98
Fekete Zoltan - 99 (part of study)
Hi! Just want to make a correction on 33. While it’s true that Ganguly helped this composition go viral by sharing the original Reddit post on Twitter, the composition was actually created by Ben Schultz. The reason I know this is because I am Ben Schultz. ;)

Proof:

Playlist of videos: